The long-term objective of this research is to develop and test automated cell image analysis for the improved diagnosis and prognosis of cervical cancer. our previous research has shown that contextual analysis (low resolution analysis of scenes) and marker analysis (morphometric changes in "normal" looking cells) adds significant information to high resolution individual cell analysis of breast, prostate, and cervical neoplasia. The previous studies were performed in several steps, which was inefficient and only permitted us to collect small data sets. The primary objective of the proposed research is to develop an Integrated Analysis System (IAS) that will employ more sophisticated methods to combine the single cell, contextual, and marker analyses. The first aim of this research is to develop a system to permit simultaneous contextual, marker, and individual cell analyses of monolayer preparations of exfoliated cervical cells. This will include the use of adaptive methodologies that incorporate a priori information to guide the analysis process. The second aim is to test the combined methodolology to determine whether it provides an improvement over any single analysis method. The third aim is to apply these technique to investigate the feasibility of discriminating persistent/progressive from regressive and negative cases. This research should have significant impact on cost-containment and availability of cervical cytology for pre-screening of cervical cancer Automated cytology can partly offset the growing shortage of qualified cytotechnologists. It can also provide more objective criteria for diagnosis and prognosis of cervical neoplasia at all levels of experience and training. The same methodology can eventually be applied to expert systems for diagnosis and teaching. Moreover, advances made in this research should be readily transferable to other types of human cancer. Thus, we expect that this research will have far-reaching effects on practical as well as scientific aspects of cancer prevention and control.